CN1645054A - Digital distorting model generating method for compensating image distortion of camera measurement - Google Patents

Abstract

A method for generating digital distortion model includes setting up and shooting control field, selecting four corner marks and making their distortion zero, leading projective difference in for realizing true 2-D control field transform, measuring out actual position and calculating out theoretical position of each control mark and confirming total distortion of each control mark, obtaining total distortion for each picture element to generate the model by interpolating based on total distortion of four marks nearest to picture element to be measured.

Description

Compensate the generation method of the Digital Distortion Model for all of photogrammetric image distortion

Technical field

The invention belongs to the Photogrammetry and Remote Sensing technical field, relate to the Digital Distortion Model for all that compensates photogrammetric image distortion and the generation method of relevant parameter (elements of interior orientation).

Background technology

Compensation to image distortion has two kinds of strategies, i.e. the Digital Distortion Model for all strategy that Guan Yong function model strategy and the present invention proposes.

Even to this day,,, adopt certain polynomial function model both at home and abroad to the compensation of all kinds of distortion of video camera almost homogeneously.This type of polynomial coefficient or have tangible physical significance, or have the physical significance of mimetism, or rely on long-term optimal design, to determine polynomial profile and corresponding coefficient thereof.

Has tangible physical significance person, for example the optics coefficient of radial distortion (k in the formula (1) ₁, k ₂) and optics tangential distortion coefficient (p ₁, p ₂):

$\left.\begin{array}{c}\mathrm{\&Delta;x}=x(k_1{r}^2+k_2{r}^4+\&CenterDot;\&CenterDot;\&CenterDot;)+p_1(r^2+2x^2)+2p_2\mathrm{xy}\\ \mathrm{\&Delta;y}=y(k_1{r}^2+k_2{r}^4+\&CenterDot;\&CenterDot;\&CenterDot;)+p_2(r^2+2y^2)+2{\mathrm{<figure-callout\; id="1"\; label="p"\; filenames="A20051001815100091.png"\; state="\{\{state\}\}">p</figure-callout>}}_1\mathrm{xy}\end{array}\right\}---\left(1\right)$

Physical significance person with mimetism, the Brown model that includes 21 parameters shown in Salmenper  model and formula (2):

$\left.\begin{array}{c}\mathrm{\&Delta;x}=a_{1}x+a_{2}y+a_3\mathrm{xy}+a_4{y}^2+a_5{x}^{2}y+a_{6}x{y}^2+\&CenterDot;\&CenterDot;\&CenterDot;+a_{21}\frac{x}{f}\\ \mathrm{\&Delta;y}=a_8\mathrm{xy}+a_9{x}^2+a_{10}{x}^{2}y+a_{11}x{y}^2+a_{12}{x}^2{y}^2+\&CenterDot;\&CenterDot;\&CenterDot;a_{21}\frac{y}{f}\end{array}\right\}---\left(2\right)$

Do not have tangible physical significance, but rely on long-term optimal design, to determine polynomial profile.This type of polynomial kind is a lot, cubic polynomial, the cubic polynomial of Schut, Gotthardt as Kupfer, Grean, function model of the photogrammetric and remote sensing that the polynomial expression that people such as Mauelshagen and Heikkil  use, the orthogonal polynomial with certain geometric meaning of Ebner and other profiles are more complicated or the like.

When using function model, follow-up data photogrammetric and remote sensing is handled, and (sometimes or even fail to agree) the tens of kinds of function models that must propose in the face of the west personage make a choice, and its consumption worker consuming time is difficult to add up.In addition, use traditional function model, must understand the imaging mechanism of imaging system, must analyze possible error source and influence degree thereof one by one and at large.Generally speaking, " versatility " that does not possess the camera that is suitable for different imaging mechanisms based on the method for ' function model '; To a great extent, do not possess compensation ability to distortion that can not mathematical simulation yet; To a great extent, must bear " the stable not to the utmost influence " determined in the multinomial coefficient again; And, use traditional function model, lack the objective standard of estimating achievement mostly; The Digital Distortion Model for all of using the present invention to propose then can be estimated achievement objectively by controlling filed.

Summary of the invention

Problem to be solved by this invention provides a kind of generation method that compensates the Digital Distortion Model for all of photogrammetric image distortion, this method not only can be easily to the image distortion compensation of the camera of different imaging mechanisms, and can obtain corresponding elements of interior orientation, with realize continue after three-dimensional measurement.

Technical scheme provided by the invention is: the generation method that a kind of Digital Distortion Model for all that compensates photogrammetric image distortion is taken the photograph may further comprise the steps:

One, sets up controlling filed;

Two, take controlling filed with the calibrating camera; Measure the image elements of exterior orientation;

Three, measure the physical location of each controlled flag on image;

Four, select four corner signs of film size, make its abnormal vanishing; Determine fit Plane, and these four corner signs are introduced height displacement by these four corner signs; All controlled flag in the film size are introduced height displacement, realize the conversion of controlling filed to true 2 d control network;

Five, according to above-mentioned four corner signs, press the 2 d dlt principle, separate and ask 8 conversion coefficients, calculate the theoretical position of each controlled flag on image in the film size;

Six,, determine the resultant distortion (f of each controlled flag on the image by the physical location and the theoretical position of each controlled flag on image ₁, f ₂, f ₃, f ₄);

Seven, obtain the resultant distortion of each pixel on the image: according to the resultant distortion of four signs that close on most with pixel to be measured on the image, by following any distribution number strong point relational expression, interpolation is obtained the resultant distortion f of this pixel _{(x, y)}, until the resultant distortion that obtains each pixel on the image; Thereby constitute the Digital Distortion Model for all of this image;

$y_{\left(\mathrm{1,2}\right)}=\frac{(y_2-y_1)}{(x_2-x_1)}(x_{\left(\mathrm{1,2}\right)}-x_1)+y_1$ $y_{\left(\mathrm{3,4}\right)}=\frac{(y_4-y_3)}{(x_4-x_3)}(x_{\left(\mathrm{3,4}\right)}-x_3)+y_3$

$f_{\left(\mathrm{1,2}\right)}=f_1+\frac{\sqrt{{(x_{\left(\mathrm{1,2}\right)}-x_1)}^2+{(y_{\left(\mathrm{1,2}\right)}-y_1)}^2}}{\sqrt{{(x_2-x_1)}^2+{(y_2-y_1)}^2}}(f_2-f_1)$

$f_{(3,4)}=f_3+\frac{\sqrt{{(x_{(3,4)}-x_3)}^2+{(y_{(3,4)}-y_3)}^2}}{\sqrt{{(x_4-x_3)}^2+{(y_4-y_3)}^2}}(f_4-f_3)$

$f_{(x,y)}=f_{\left(\mathrm{1,2}\right)}+\frac{\sqrt{{(x-x_{\left(\mathrm{1,2}\right)})}^2+{(y-y_{\left(\mathrm{1,2}\right)})}^2}}{\sqrt{{(x_{\left(\mathrm{3,4}\right)}-x_{\left(\mathrm{1,2}\right)})}^2+{(y_{\left(\mathrm{3,4}\right)}-y_{\left(\mathrm{1,2}\right)})}^2}}(f_{\left(\mathrm{3,4}\right)}-f_{\left(\mathrm{1,2}\right)})$

X, y are horizontal stroke, the ordinate of pixel to be measured in the formula; (x ₁, y ₁), (x ₂, y ₂), (x ₃, y ₃), (x ₄, y ₄) be respectively horizontal stroke, the ordinate of four signs that pixel to be measured closes on most; x _(1,2), y _(1,2)Be (x ₁, y ₁) and (x ₂, y ₂) line on horizontal stroke, the ordinate of point; x _(3,4), y _(3,4)Be (x ₃, y ₃) and (x ₄, y ₄) line on horizontal stroke, the ordinate of point; X wherein _(1,2)=x, x _(3,4)=x.

For ease of realizing three-dimensional measurement to photography target, by the distortion distribution principle of the abnormal vanishing of four corner signs, according on the horizontal photo with the cornerwise corresponding line segment of four corner signs of controlling filed, measure corresponding elements of interior orientation.

Digital Distortion Model for all of the present invention can be based upon on the basis of accurate 2 d control network; Also can set up very two dimension or three-dimensional controlling filed according to the needs of self-technique condition and actual application environment.Mm in full ²Master grating, can be considered to true 2 d control network.For satisfying the open-air controlling filed of aviation image needs, can set up three-dimensional controlling filed.

The present invention not only can be easily to the image distortion compensation of the camera of different imaging mechanisms, and can obtain corresponding elements of interior orientation, with realize continue after three-dimensional measurement.Its outstanding advantage comprises his preciseness, simplification, extensive adaptability and to the inspectability of computer hardware, totally four aspects.

A. preciseness

The foundation of Digital Distortion Model for all makes photogrammetric process be in more rigorous environment.Like this, perhaps improve photogrammetric performance and quality, perhaps improved the precision and the efficient of photogrammetric process.Enforcement of the present invention because of the intensive marker group that can discern automatically is provided at object space, is estimated photogrammetric result scrupulously so can be used to very objective.

B. simplification

Generate Digital Distortion Model for all and the overall process that corresponding elements of interior orientation is provided, can take the technology path of " will not analyze " to reason and size that the imaging system imaging process produces all distortion, directly measure the resultant distortion that a variety of causes on each pixel causes distortion, avoided the complicated and selection problem of polynomial expression profile not to the utmost reliably in ' function model ' strategy fully.

C. extensive adaptability

The present invention can extensively adapt to multiple imaging system Digital Distortion Model for all foundation and the overall process of corresponding elements of interior orientation is provided.These imaging systems comprise: all kinds of solid-state cameras (Solid State Camera), stereocamera, kinematograph, cinetheodollite, high-speed camera, military reconnaissance camera, gun camera, underwater camera, other pairs medium and multimedium video camera, x X-ray machine X, optical microscope picture pick-up device, electron microscope picture pick-up device, fish-eye lens video camera, Fundus photography machine, ballistic camera or the like also comprise other ground and aerial metric camera, graticule mesh metric camera and half metric camera that uses.

D. to the inspectability of computer hardware

Theory of the present invention and method can be applicable to the check of the various imaging computer hardwares of domestic and imported.For example, to the check of electron microscope picture pick-up device deformation of image, to check of ballistic camera, military reconnaissance camera, high-speed camera, underwater camera and graticule mesh metric camera etc.

Description of drawings

The Digital Distortion Model for all that Fig. 1 sets up for certain fisheye camera;

Fig. 2 is the indoor accurate 2 d control network of sensor information engineering college of Wuhan University;

Fig. 3 is the present invention interpolation theory figure that distorts;

Fig. 4 is the original flake image of certain buildings;

Fig. 5 is to the image after the image process Digital Distortion Model for all correction of Fig. 4;

Fig. 6 is to the image after the image process conventional processing of Fig. 5;

Fig. 7 is the original flake image on Zhonghuan, Hongkong square;

Fig. 8 is to the image after the image process Digital Distortion Model for all correction of Fig. 7;

Fig. 9 is to use the facade striograph of shooting of flake digital camera and Jiang-Han Area, the Wuhan City road buildings after the present invention handles.

Embodiment

Digital Distortion Model for all DDM of the present invention (Digital Distortion Model) is a kind of 3-D geometric model that ' fluctuating ' arranged.Its ' planimetric coordinates ' is the numbering of capable battle array of digitized video and array, and its ' elevation ' is the overall distortion of respective pixel.Establish Digital Distortion Model for all at flake digital camera photographic image, as shown in Figure 1.

In essence, Digital Distortion Model for all is the set of the resultant distortion of each pixel on the image.

For nearly all imaging system is set up Digital Distortion Model for all, purpose is the image that obtains without any distortion, and measures corresponding elements of interior orientation simultaneously, with realize continue after three-dimensional measurement.Digital Distortion Model for all is different from the function model of domestic and international the past in essence fully.

(1) foundation of controlling filed

The generation of Digital Distortion Model for all relies on the foundation of certain controlling filed, comprises the foundation of 2 d control network, accurate 2 d control network or three-dimensional controlling filed.The present invention can be as required, at different environments for use, set up area suitable have enough controlling fileds of intensive sign.As be applicable to microscopical several mm ²Grating, be applicable to tens of km of aviation image ²Open-air controlling filed.Indoor several m ²The foundation of controlling filed can be with reference to textbook " close-range photogrammetry " (publishing house of Wuhan University, 2002 publishes); Outdoor several km ²The foundation of controlling filed can be with reference to monograph " commercial measurement " (publishing house of Wuhan University, in October, 2004 publish) and " aerophotogrammetric field work standard ", to measure the three dimensional space coordinate of intensive sign.

Indoor accurate 2 d control network is a kind of basic work of setting up multiple close shot imaging system Digital Distortion Model for all.Sensor information engineering college of Wuhan University establishes indoor accurate 2 d control network (as shown in Figure 2), and this controlling filed area is that (3.75m * 2.50m), contain 1 350 light echo reflection controlled flag, the precision of its three-dimensional coordinate all directions is about ± 0.20mm.

Assist by us, be based upon the indoor accurate 2 d control network of The Hong Kong Polytechnic University industrial center.This controlling filed area is that (3.75m * 2.50m), contain 1 944 light echo reflection controlled flag, the precision of its three-dimensional coordinate all directions is about ± 0.15mm.

The controlling filed of this kind scale, the world sets up first.It is applicable to the foundation of the Digital Distortion Model for all of all kinds of generic digital camera, also is applicable to limited focalizing and the focalize foundation of Digital Distortion Model for all of video camera of infinite distance.

(2) foundation of flake digital camera Digital Distortion Model for all and application

Focalize by given photo distance and practicality,, set up Digital Distortion Model for all (DDM) the flake digital camera (field angle is respectively 120 ° and nearly 180 °) of two kinds of models.It sets up process:

1. focalize by practicality, use the flake digital camera of two kinds of models, take aforesaid indoor accurate 2 d control network, and measure the outer orientation vertical element of taking the photograph website;

2. measure the physical location of each controlled flag on image;

3. select four corner signs of film size, make its abnormal vanishing;

4. determine fit Plane, and these four corner signs are introduced height displacement by these four corner signs; All controlled flag in the film size are introduced height displacement, realize of the conversion of accurate 2 d control network to virtual true 2 d control network;

5. according to above-mentioned four corner signs, press 2 d dlt principle (referring to above-mentioned textbook " close-range photogrammetry "), separate and ask 8 conversion coefficients, calculate the theoretical position of each controlled flag on image in the film size;

6. by physical location and the theoretical position of each controlled flag on image, determine the resultant distortion (f of each controlled flag on the image ₁, f ₂, f ₃, f ₄);

7. obtain the resultant distortion of each pixel on the image: as shown in Figure 3, according to the resultant distortion of four monumented points that close on most with pixel to be measured on the image (1,2,3,4), by following any distribution number strong point relational expression, interpolation is obtained the resultant distortion f of this pixel _{(x, y)}, until the resultant distortion that obtains each pixel on the image; Thereby constitute the Digital Distortion Model for all of flake digital camera photographic image;

x _(1，2)＝x； $y_{\left(\mathrm{1,2}\right)}=\frac{(y_2-y_1)}{(x_2-x_1)}(x_{\left(\mathrm{1,2}\right)}-x_1)+{\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="A20051001815100091.png"\; state="\{\{state\}\}">y</figure-callout>}}_1$

x _(3，4)＝x； $y_{\left(\mathrm{3,4}\right)}=\frac{(y_4-y_3)}{(x_4-x_3)}(x_{\left(\mathrm{3,4}\right)}-x_3)+y_3$

$f_{\left(\mathrm{1,2}\right)}={\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A20051001815100091.png"\; state="\{\{state\}\}">f</figure-callout>}}_1+\frac{\sqrt{{(x_{\left(\mathrm{1,2}\right)}-x_1)}^2+{(y_{\left(\mathrm{1,2}\right)}-y_1)}^2}}{\sqrt{{(x_2-x_1)}^2+{(y_2-y_1)}^2}}(f_2-f_1)$

$f_{(3,4)}=f_3+\frac{\sqrt{{(x_{(3,4)}-x_3)}^2+{(y_{(3,4)}-y_3)}^2}}{\sqrt{{(x_4-x_3)}^2+{(y_4-y_3)}^2}}(f_4-f_3)$

$f_{(x,y)}=f_{\left(\mathrm{1,2}\right)}+\frac{\sqrt{{(x-x_{\left(\mathrm{1,2}\right)})}^2+{(y-y_{\left(\mathrm{1,2}\right)})}^2}}{\sqrt{{(x_{\left(\mathrm{3,4}\right)}-x_{\left(\mathrm{1,2}\right)})}^2+{(y_{\left(\mathrm{3,4}\right)}-y_{\left(\mathrm{1,2}\right)})}^2}}(f_{\left(\mathrm{3,4}\right)}-f_{\left(\mathrm{1,2}\right)})$

8. according to the Digital Distortion Model for all related data, measure the elements of interior orientation of revising distortion back image.

9. by known Digital Distortion Model for all and corresponding elements of interior orientation, implement Flame Image Process and Photogrammetric Processing.

(3) flake digital camera Digital Distortion Model for all is used example

Example 1: take high-rise with this flake digital camera, its raw video, through revised image and the image after routine is handled of distorting shows respectively as Fig. 4, Fig. 5 and Fig. 6.Handled overall process only several minutes consuming time.The ground of comparing uses conventional digital camera, can not obtain identical achievement once.

Example 2: the upright face picture of buildings after the original flake image of Zhonghuan, Hongkong square (Center Plaza) and use above-mentioned steps of the present invention are handled, show respectively as Fig. 7 and Fig. 8.The ground of comparing uses conventional digital camera, can not obtain identical achievement once.

Example 3: use the flake digital camera to take the photograph 4 width of cloth images, use above-mentioned steps of the present invention to handle, made up the l of Jiang-Han Area, Wuhan City road buildings: 200 large scale facade striograph, as shown in Figure 9.The ground of comparing uses conventional digital camera, can not obtain identical achievement once.

Among the present invention, can adopt three-dimensional controlling filed to replace above-mentioned accurate 2 d control network, can obtain identical result.And, when adopting true 2 d control network, can directly implement above-mentioned the and 5. go on foot, and obtain identical result up to the 9. step.

Claims (3)

1. compensate the generation method of the Digital Distortion Model for all of photogrammetric image distortion, it is characterized in that: may further comprise the steps: 1. set up controlling filed; 2. take controlling filed with the calibrating camera; Measure the image elements of exterior orientation; 3. measure the physical location of each controlled flag on image; 4. select four corner signs of film size, make its abnormal vanishing; On controlling filed, determine fit Plane, and these four corner signs are introduced height displacement by these four corner signs; All controlled flag in the film size are introduced height displacement, realize the conversion of controlling filed to true 2 d control network; 5. according to above-mentioned four corner signs, press the 2 d dlt principle, separate and ask 8 conversion coefficients, calculate the theoretical position of each controlled flag on image in the film size; 6. by physical location and the theoretical position of each controlled flag on image, determine the resultant distortion (f of each controlled flag on the image ₁, f ₂, f ₃, f ₄); 7. obtain the resultant distortion of each pixel on the image: according to the resultant distortion of four signs that close on most with pixel to be measured on the image, by following any distribution number strong point relational expression, interpolation is obtained the resultant distortion f of this pixel _{(x, y)}, until the resultant distortion that obtains each pixel on the image; Thereby constitute the Digital Distortion Model for all of this image;

$y_{\left(\mathrm{1,2}\right)}=\frac{(y_2-y_1)}{(x_2-x_1)}(x_{\left(\mathrm{1,2}\right)}-x_1)+y_1$ $y_{\left(\mathrm{3,4}\right)}=\frac{(y_4-y_3)}{(x_4-x_3)}(x_{\left(\mathrm{3,4}\right)}-x_3)+y_3$

$f_{\left(\mathrm{1,2}\right)}=f_1+\frac{\sqrt{{(x_{\left(\mathrm{1,2}\right)}-x_1)}^2+{(y_{\left(\mathrm{1,2}\right)}-y_1)}^2}}{\sqrt{{(x_2-x_1)}^2+{(y_2-y_1)}^2}}(f_2-f_1)$

$f_{\left(\mathrm{3,4}\right)}=f_3+\frac{\sqrt{{(x_{\left(\mathrm{3,4}\right)}-x_3)}^2+{(y_{\left(\mathrm{3,4}\right)}-y_3)}^2}}{\sqrt{{(x_4-x_3)}^2+{(y_4-y_3)}^2}}(f_4-f_3)$

$f_{(x,y)}=f_{\left(\mathrm{1,2}\right)}+\frac{\sqrt{{(x-x_{\left(\mathrm{1,2}\right)})}^2+{(y-y_{\left(\mathrm{1,2}\right)})}^2}}{\sqrt{{(x_{\left(\mathrm{3,4}\right)}-x_{\left(\mathrm{1,2}\right)})}^2+{(y_{\left(\mathrm{3,4}\right)}-y_{\left(\mathrm{1,2}\right)})}^2}}(f_{\left(\mathrm{3,4}\right)}-f_{\left(\mathrm{1,2}\right)})$

X, y are horizontal stroke, the ordinate of pixel to be measured in the formula; (x ₁, y ₁), (x ₂, y ₂), (x ₃, y ₃), (x ₄, y ₄) be respectively horizontal stroke, the ordinate of four signs that pixel to be measured closes on most; x _(1,2), y _(1,2)Be (x ₁, y ₁) and (x ₂, y ₂) line on horizontal stroke, the ordinate of point; x _(3,4), y _(3,4)Be (x ₃, y ₃) and (x ₄, y ₄) line on horizontal stroke, the ordinate of point; X wherein _(1,2)=x, x _(3,4)=x.

2. method according to claim 1 is characterized in that: by the distortion distribution principle of the abnormal vanishing of four corner signs, according on the horizontal photo with the cornerwise corresponding line segment of four corner signs of controlling filed, measure corresponding elements of interior orientation.

3. according to claim 1,2 described methods, it is characterized in that: controlling filed be as the criterion 2 d control network, true 2 d control network or three-dimensional controlling filed.

Images